Support log-level for controller logs

[dorroddorrod]

@jgwest EDIT:
Enhancement Request:
Add support for configurable log levels via parameter, eg the ability to set additional log levels beyond just the --debug parameter support we have now. For example, as requested, ability to set the log level to only show error or warn.

Original Text from @dorroddorrod
Hi,
Is there any way to control the log level for the application-set controller ?

[jgwest]

Hi @dorroddorrod , you can add --debug parameter to the argocd-applicationset-controller Deployment, eg:
kubectl edit deployment/argocd-applicationset-controller, and add this line:

spec:
  template:
    spec:
      containers:
      - command:
        - applicationset-controller
        - --debug  # <-------- add this

Or you can edit the install.yaml, add the line as above, and reapply using kubectl apply -f install.yaml:
https://github.com/argoproj-labs/applicationset/blob/48a7b47da0552b9e5a518704e25489ab4a36645c/manifests/install.yaml#L3655

[dorroddorrod]

Hi @jgwest
Thanks for your reply
I want to change the log level to warn and not to debug.
Currently the info logs are bombarding our logging database (ES)

[jgwest]

Ah, gotcha, sorry about that 😅, we don't have that currently, and the logs are definitely verbose by default. Let me convert this issue into an enhancement for that.

[rumstead]

Do we have any thoughts on how this should be implemented? Ie, instead of taking debug argument, accept an arbitrary log level?

Accept 2 for backward compatibility?
Ie, --debug == --loglevel debug
--loglevel info, trace, debug, error

Any suggestions on which info logs should be changed to debug/trace?

I am interested in completing this issue :)

[jgwest]

Awesome, thanks @rumstead.

Exactly as you said, it couldn't hurt to keep --debug for the time being, and add --loglevel (info, trace, debug, error) (just like Argo CD)

Open to suggestions re: which info logs should be downgraded, if they are not providing any value.

[rumstead]

@jgwest feel free to assign to me

[rumstead]

Ok, so the majority of the logs are coming from the git clients in ArgoCD.

The git client every time it runs a command -> hits this method -> which captures tracing spans

Since it uses logrus.New(), it gets the default info logs. I am not sure if there is a way to configure this.

@jgwest any comments or suggestions here? You will see very similar logs in the repo-server of ArgoCD.

time="2021-05-27T15:50:30Z" level=info msg="git fetch origin --tags --force" dir="/tmp/my_internal_git_repo" execID=KIlVw
time="2021-05-27T15:50:32Z" level=info msg=Trace args="[git fetch origin --tags --force]" dir="/tmp/my_internal_git_repo" operation_name="exec git" time_ms=1769.0971630000001
time="2021-05-27T15:50:32Z" level=info msg="git checkout --force git_sha" dir="/tmp/my_internal_git_repo" execID=AmjdR
time="2021-05-27T15:50:32Z" level=info msg=Trace args="[git checkout --force git_sha]" dir="/tmp/my_internal_git_repo" operation_name="exec git" time_ms=4.057425
time="2021-05-27T15:50:32Z" level=info msg="git clean -fdx" dir="/tmp/my_internal_git_repo" execID=x3uMq
time="2021-05-27T15:50:32Z" level=info msg=Trace args="[git clean -fdx]" dir="/tmp/my_internal_git_repo" operation_name="exec git" time_ms=2.401413
time="2021-05-27T15:50:32Z" level=info msg="git ls-files --full-name -z -- apps/**/config.json" dir="/tmp/my_internal_git_repo" execID=zeesc
time="2021-05-27T15:50:32Z" level=info msg=Trace args="[git ls-files --full-name -z -- apps/**/config.json]" dir="/tmp/my_internal_git_repo" operation_name="exec git" time_ms=1.9009090000000002
time="2021-05-27T15:50:32Z" level=info msg="git fetch origin --tags --force" dir="/tmp/my_internal_git_repo" execID=zsqKe
time="2021-05-27T15:50:33Z" level=info msg=Trace args="[git fetch origin --tags --force]" dir="/tmp/my_internal_git_repo" operation_name="exec git" time_ms=390.86078499999996
time="2021-05-27T15:50:33Z" level=info msg="git checkout --force git_sha" dir="/tmp/my_internal_git_repo" execID=Ck9ui
time="2021-05-27T15:50:33Z" level=info msg=Trace args="[git checkout --force git_sha]" dir="/tmp/my_internal_git_repo" operation_name="exec git" time_ms=4.16638
time="2021-05-27T15:50:33Z" level=info msg="git clean -fdx" dir="/tmp/my_internal_git_repo" execID=SieSR
time="2021-05-27T15:50:33Z" level=info msg=Trace args="[git clean -fdx]" dir="/tmp/my_internal_git_repo" operation_name="exec git" time_ms=2.6018920000000003
time="2021-05-27T15:50:33Z" level=info msg="git fetch origin --tags --force" dir="/tmp/my_internal_git_repo" execID=CZjtN
time="2021-05-27T15:50:35Z" level=info msg=Trace args="[git fetch origin --tags --force]" dir="/tmp/my_internal_git_repo" operation_name="exec git" time_ms=1648.057597
time="2021-05-27T15:50:35Z" level=info msg="git checkout --force git_sha" dir="/tmp/my_internal_git_repo" execID=X7MUV
time="2021-05-27T15:50:35Z" level=info msg=Trace args="[git checkout --force git_sha]" dir="/tmp/my_internal_git_repo" operation_name="exec git" time_ms=4.729625
time="2021-05-27T15:50:35Z" level=info msg="git clean -fdx" dir="/tmp/my_internal_git_repo" execID=QkYjC
time="2021-05-27T15:50:35Z" level=info msg=Trace args="[git clean -fdx]" dir="/tmp/my_internal_git_repo" operation_name="exec git" time_ms=2.3974349999999998

[jgwest]

@rumstead, looks like we're stuck with those messages for now, without an upstream Argo CD patch to reduce their verbosity.

[rumstead]

Ok, the PR adds the scaffolding to support multiple log levels.